Abstract
Remote platform-based laser positioning applications are severely affected by onboard vibration. To eliminate the effect, sensing and compensating the vibration by equal but opposite actuation is necessary. However, existing methods possess various limitations due to the technical challenges as follows: 1) the vibration has multiple time-variant dominant frequencies; 2) the bandwidth is broad; 3) there is a significant phase shift in signals caused by inherent delay; 4) the sensor (i.e., gyroscope) generates noise and integration drift disturbances. To overcome these challenges, we propose a novel sensing scheme in this paper, applying the Taylor series fore-casting and the Recursive Least Square (RLS)-based filter. Both techniques are designed to process signals with features listed in challenges 1) and 2). The Taylor series forecasting is a time series forecasting technique that eliminates the delay-induced phase shift with higher accuracy than existing methods. The RLS-based filter is an adaptive filter that applies a regression model to the input and achieves frequency domain separation with negligible phase shift. In this paper, both offline and online real-time (RT) experiments are conducted for validation, using the vibration signal sampled from a truck-based system. The proposed method shows higher accuracy than existing methods and achieves compensation rates over 70%.